Fitness Trackers: What to Track and What Not To
Dr. Sahanawaz, PhD, CPT
Your smartwatch is not lying to you on purpose. It's just guessing more than the marketing page admits.
Fitness bands, rings, and watches have never been more popular. Anyone who goes for a walk now wants a step-tracking band or ring on their wrist. More enthusiastic users want heart rate, SpO2, sleep, stress, and recovery tracked too — they don't want a single health parameter left unmeasured. And modern wearables do claim to track all of this.
The real question is: how good are they at it, and do you actually need to track all of it?
Quick answer, if you're skimming:
Steps — good enough for trend-tracking; don't overpay for accuracy
Calories burned — never trust this number; error runs up to 44%
Heart rate — genuinely reliable at rest and for trend during training
HRV / "stress" score — noisy, brand-dependent, not clinically equivalent
SpO2 — fine as a casual check; not for medical decisions
Sleep staging — best devices hit 65–75% agreement with a sleep lab; some hit barely 26%
Stress & recovery scores — the least validated metric of all, and can create anxiety that wasn't there before
The full breakdown, with the studies behind each number, is below.
Steps: track the trend, not the number
The most popular metric people chase is daily step count, largely because of the popular belief that 10,000 steps a day improves longevity and quality of life. Whether that specific number is the "right" one is a separate discussion — but higher daily physical activity is, in general, good for health, so aiming for a high step count is a reasonable goal.
Step-tracking accuracy varies by device, and premium trackers are often closer to the true count than cheap generic bands. But that doesn't mean you need an expensive watch to track steps "more accurately." Even premium devices aren't perfectly accurate. Across 20+ peer-reviewed validation studies on Garmin, Apple Watch, Fitbit, Oura, WHOOP, and Samsung devices, accuracy depends heavily on walking speed, arm swing, and device placement — often more than brand or price does. Slow walking produces roughly eight times more error than brisk walking, and free-living accuracy tends to run two to three times worse than accuracy measured in a controlled lab. A large multi-review analysis spanning 249 studies and over 430,000 participants confirms this pattern of brand-independent variability [O'Driscoll et al., 2024].
The purpose of a step tracker isn't to give you a precise number — it's to remind you: "you need 2,000 more steps to hit your goal today." Whether your watch says 9,999 or a cheap band says 9,500 when you actually walked 10,000, your health outcome is identical. There's no point spending extra money purely to shave a few percentage points off step-count error.
Calories burned: don't trust any device
No wearable, regardless of price, is good at estimating calories burned. This is consistently the weakest metric across every validation study. One comparative study found wearables were reasonably accurate for heart rate, steps, distance, and sleep duration (average error around 10%), while energy expenditure error ran as high as 44%. That is not a rounding error — a device could tell you that you burned 500 kcal when you actually burned 280, or vice versa. Never make eating or training decisions based on the "calories burned" number on your wrist.
Heart rate and SpO2: fine for trend, not for medical decisions
Resting heart rate tracking has genuinely improved. In steady-state or resting conditions, top wearables are quite close to ECG references. In one 24-hour validation study, the Apple Watch showed a mean absolute error of under 6% compared to an ambulatory ECG, and newer devices like Oura have shown even tighter agreement for nighttime resting heart rate with a mean error under 2% against an ECG reference in one 2025 study.
Heart rate variability (HRV) — often marketed as a "stress" or "recovery" score — is a different story. A study validating the Apple Watch Series 9 and Ultra 2 against a chest-strap ECG reference found HRV was underestimated on average, with almost 29% mean error, and the results did not meet the pre-set threshold for clinical equivalence. Accuracy also varies enormously by brand: the same 2025 study comparing five wearables against ECG found Oura rings showed the strongest agreement with the reference device, while Garmin and Polar devices showed the weakest.
For SpO2 (blood oxygen), consumer wrist devices are convenience tools, not diagnostic ones. A study comparing several consumer smartwatches to a clinical-grade pulse oximeter found the best-performing device (Apple Watch Series 7) had a mean absolute error of about 2.2%, while accuracy differed meaningfully between devices. Accuracy can also vary by skin tone: a large hospital-based study found pulse oximeter readings ran 0.6–1.5 percentage points higher for patients with darker skin tones, and the rate of falsely reassuring "normal" readings when true oxygen levels were actually low was several times higher in darker-skinned patients. This bias has been documented across multiple independent reviews, not just one study.
Bottom line: use heart rate and SpO2 numbers for a general trend during exercise ("how elevated does my heart rate get during intervals?"). Don't rely on them for medical decisions — for that, use a validated medical-grade device.
Sleep, stress, and recovery: the most-marketed, least-reliable trio
This is where the biggest gap exists between marketing and reality. Sleep, stress, and recovery scores are sold as "premium" or "core" features on nearly every modern wearable. Two questions matter here: (1) can these devices actually measure these things accurately, and (2) do you even need them measured by a device at all?
On accuracy: True sleep staging (light, deep, REM) requires measuring brain waves via EEG — polysomnography, the clinical gold standard. Wrist and ring devices instead infer sleep stages from motion and pulse-wave patterns using an algorithm, which is a fundamentally different and less direct method.
The data confirms this gap. A 2023 multicenter study comparing 11 consumer sleep trackers against in-lab polysomnography across nearly 350,000 sleep epochs found substantial performance variation between devices, with agreement scores (macro F1) ranging from as low as 0.26 to as high as 0.69 — meaning some devices were barely better than a coin flip at correctly staging sleep, while the best ones were only moderately reliable. A separate 2024 meta-analysis pooling multiple studies concluded that total sleep time, sleep efficiency, sleep latency, and time awake after sleep onset all differed significantly from polysomnography measurements, and that consumer wrist trackers are not yet considered valid instruments compared to the clinical gold standard. Device type matters too: accelerometer-only devices tend to have the lowest staging accuracy, around 65%, while devices combining motion with pulse-wave sensors reach roughly 65–75% accuracy for sleep stages and about 90% for simply distinguishing sleep from wake.
Stress and recovery scores are even further removed from anything directly measurable — they're algorithmic estimates built on HRV, resting heart rate, and sleep data, each of which already carries its own error margin. Stacking an uncertain metric (stress) on top of another uncertain metric (HRV) compounds the unreliability rather than cancelling it out.
On necessity: You don't need a third-party device to tell you whether you slept well. If your sleep-wake time is fairly fixed, you wake up without needing an alarm most days, and you don't feel drowsy through the day, your sleep is very likely already adequate — no tracking required. If you can't wake up without an alarm and feel groggy all day, the fix is to work on sleep hygiene and circadian consistency, not to buy a device that quantifies the problem you already know you have. If you genuinely need precise data — for example, suspected sleep apnea — the right move is a clinical sleep study, not a consumer ring.
Stress scores carry a specific risk: a device telling you "you are in high stress" when you are not can itself become a stressor. That false signal can create anxiety and disrupt sleep — the wearable ends up causing the exact problem it claims to monitor.
Summary table: what wearables are actually good for
Metric | Typical error vs. gold standard | Verdict |
|---|---|---|
Step count | Roughly 3–10% in controlled walking; 2–3x worse in daily life | Good enough for trend-tracking; premium devices not worth it for this alone |
Calories burned | Up to ~44% error | Never rely on this for diet or training decisions |
Resting heart rate | Roughly 2–6% error on better devices | Reasonably reliable for trend and training zones |
Heart rate variability | ~8–28% error, varies widely by brand | Treat as a rough trend at best; not clinically equivalent |
SpO2 | ~2–3% error on better devices; degrades with darker skin tones | Fine as a casual check; not for medical decisions |
Sleep staging | 65–75% agreement with polysomnography (best devices) | Useful for consistency trends only, not precise sleep architecture |
Stress/recovery score | No consistent validation against a physiological gold standard | Least reliable metric category; can create false alarms and unnecessary anxiety |
The practical takeaway
Wearables are good at showing you direction and trend — are you moving more this week than last, is your resting heart rate drifting up, are you roughly hitting your step goal? They are not good at giving you precise numbers, and they are especially unreliable for calories, HRV-based stress scores, and detailed sleep staging.
So: track steps loosely, ignore the calorie counter completely, use heart rate and SpO2 as rough trend indicators rather than medical data, and for sleep, stress, and recovery — trust how you actually feel over what an algorithm on your wrist tells you. If something feels genuinely off, see a qualified healthcare provider and get proper diagnostics, rather than outsourcing that judgment to a 50k sensor.
If this was useful, subscribe. This newsletter breaks down fitness, health, and recovery claims the same way — checking what the research actually says before deciding what's worth your time, money, or attention.
Dr. Sahanawaz holds a PhD in Organic Chemistry and is a certified personal trainer. He writes about evidence-based fitness systems for people who don't have unlimited time, willpower, or budget to spend on their health.
References
O'Driscoll R, et al. (2024). Wearable technology and physical activity: a systematic review of 24 reviews covering 249 studies. Sports Medicine. DOI: 10.1007/s40279-024-02077-2
Choe & Kang (2025). Meta-analysis of step-count accuracy across consumer wearable brands.
Reddy RK, et al. (2018). Comparative accuracy of wearable devices for measuring steps, heart rate, and energy expenditure. PMC.
Fuller D, et al. (2020). Reliability and validity of commercially available wearable devices for measuring steps, energy expenditure, and heart rate: systematic review. JMIR mHealth and uHealth, 8(9):e18694. DOI: 10.2196/18694
Lee H, et al. (2023). Accuracy of 11 wearable, nearable, and airable consumer sleep trackers: prospective multicenter validation study. JMIR mHealth and uHealth, 11:e50983. DOI: 10.2196/50983
Kim JH, et al. (2024). Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis. PMC.
Evaluating accuracy in five commercial sleep-tracking devices compared to research-grade actigraphy and polysomnography (2024). PMC.
Dial B, et al. (2025). Validation of nocturnal resting heart rate and heart rate variability in consumer wearables. Physiological Reports. DOI: 10.14814/phy2.70527
Validity of Apple Watch Series 9 and Ultra 2 for serial measurements of heart rate variability and resting heart rate (2024). PMC.
Bent B, et al. (2019). Accuracy of consumer wearable heart rate measurement during an ecologically valid 24-hour period: intraindividual validation study. PMC.
Pipek LZ, et al. (2023). Investigating the accuracy of blood oxygen saturation measurements in common consumer smartwatches. PLOS Digital Health, 2(7):e0000296. DOI: 10.1371/journal.pdig.0000296
The impact of skin tone on performance of pulse oximeters used by NHS England COVID Oximetry @home scheme (2026). BMJ. DOI: 10.1136/bmj-2025-085535
Effect of skin tone on the accuracy of the estimation of arterial oxygen saturation by pulse oximetry: a systematic review (2024). ScienceDirect.
Note: Some source studies validate specific device models (e.g., Apple Watch, Oura, Fitbit, WHOOP, Garmin); figures cited here reflect the ranges reported for better-performing consumer devices in each study, not universal figures for all wearables on the market.
